Datasets:
Thanks so much!
#2
by
jkeisling
- opened
No technical questions, just wanted to say thanks so much for handling the hard wrapping. You've saved me hours!
You're welcome! Glad others found it useful.
additionally, since regex ultimately has its limitations/you can only get so far with nested rules etc, I think some sort of small model (either encoder for token classification or decoder for rewriting like jina's reader model) trained on this type of task would be the "next level" in terms of improving accuracy. Unfortunately, I haven't had time for that yet, but maybe one day..